Lift Off · My Substack
The economics and architecture of enterprise AI. What it costs. How it’s built. What I’m learning.
Latest · 5 Oct 2026
Six months ago I wrote about four AI agents and a chatbot. Since then it has become a trading desk with an org chart with 11 agents, a memory layer, and an agentic verification process.
Read on Substack ↗ New here? Start with Architecting AI ↗
More from Lift Off. TopicAll topics AI investment & finance Workflow architecture Practitioner's takes Career & path 14 Sept 2026 Salesforce spends $300 million a year on models and books $1.5 billion selling the layer that runs them.
↗ 31 Aug 2026 I passed the CCAO-F and CCA-F exams, my take on what is happening in the AI market and free resources to help you prepare.
↗ 20 Jul 2026 A well-built workflow costs 12 cents a case to run. Whether that saves you anything depends entirely on volume.
↗ 23 Jun 2026 The AI knowledge layer problem and what a working one actually looks like.
↗ 31 May 2026 Here's what changed and where AI is heading.
↗ 19 May 2026 Two announcements last month signaled the same architectural shift. Most people read them as product news. They're not.
↗ 22 Apr 2026 AI can be a lot more than a chatbot. This is what a real agentic workflow looks like when you wire it into a data-dense domain.
↗ 13 Apr 2026 A practitioner's framework for enterprise AI that compounds rather than collapses.
↗ 25 Mar 2026 Disclaimer: This space moves fast. Something in here may already be outdated by the time you read it. This is a live snapshot.
↗ 16 Mar 2026 I've spent 15 years inside the enterprise software ecosystem, implementations, commercial strategy, and watching the same cycle through boardrooms. Here's my honest read on what's actually changing.
↗ 10 earlier pieces
Stay in the loop
Get the next piece. One email when something new is published on Lift Off.